Implementing a consumer
As of 11-June-2009, the Grouper change log has a bare-bones implementation. In fact, the only notifications which are coded are for group type add/delete, and group add/update/delete, field add/update/delete, stem add/update/delete, group membership list add/update/delete. Others are soon to follow, which will complete the first pass of this new module.
To implement a consumer:
1. Make sure the change log is enabled (it is by default) in the grouper.properties
# if we should insert records into grouper_change_log_temp when events happen # defaults to true changeLog.enabled = true
2. Enable (it is by default) the grouperLoader daemon process in the grouper-loader.properties which copies change log data from the temp table to the change log table (this is what ensures data integrity, strict ordering, unique id's, etc from disparate and concurrent grouper API sources). Note, by default this runs every minute at 10 seconds to the minute, the cron can generally be left blank unless this schedule needs to be adjusted
# should the change log temp to change log daemon run? changeLog.changeLogTempToChangeLog.enable = true #quartz cron-like schedule for change log temp to change log daemon, the default is 50 seconds after every minute: 50 * * * * ? #leave blank to disable this changeLog.changeLogTempToChangeLog.quartz.cron =
3. Extend the base class for a change log consumer. Code the notification part which sends appropriate data to external systems. In this case, it just sends the data to stdout. Note the error handling here, the keeping track of which change log sequence has been successful, the processing in a batch (currently 100 records will be sent at a time), and the coding to the change log events with constants (DONT USE string01, string02, etc)
package edu.internet2.middleware.grouper.changeLog.consumer;
import java.util.List;
import edu.internet2.middleware.grouper.changeLog.ChangeLogConsumerBase;
import edu.internet2.middleware.grouper.changeLog.ChangeLogEntry;
import edu.internet2.middleware.grouper.changeLog.ChangeLogLabels;
import edu.internet2.middleware.grouper.changeLog.ChangeLogProcessorMetadata;
import edu.internet2.middleware.grouper.changeLog.ChangeLogTypeBuiltin;
/**
* just print out some of the events
*/
public class PrintTest extends ChangeLogConsumerBase {
/**
* @see edu.internet2.middleware.grouper.changeLog.ChangeLogConsumerBase#processChangeLogEntries(java.util.List, edu.internet2.middleware.grouper.changeLog.ChangeLogProcessorMetadata)
*/
@Override
public long processChangeLogEntries(List<ChangeLogEntry> changeLogEntryList,
ChangeLogProcessorMetadata changeLogProcessorMetadata) {
long currentId = -1;
//try catch so we can track that we made some progress
try {
for (ChangeLogEntry changeLogEntry : changeLogEntryList) {
currentId = changeLogEntry.getSequenceNumber();
//if this is a group type add action and category
if (changeLogEntry.equalsCategoryAndAction(ChangeLogTypeBuiltin.GROUP_TYPE_ADD)) {
//print the name from the entry
System.out.println("Group type add, name: "
+ changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_TYPE_ADD.name));
}
if (changeLogEntry.equalsCategoryAndAction(ChangeLogTypeBuiltin.GROUP_TYPE_DELETE)) {
//print the name from the entry
System.out.println("Group type delete, name: "
+ changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_TYPE_DELETE.name));
}
//if this is a group add action and category
if (changeLogEntry.equalsCategoryAndAction(ChangeLogTypeBuiltin.GROUP_ADD)) {
//print the name from the entry
System.out.println("Group add, name: "
+ changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_ADD.name));
}
if (changeLogEntry.equalsCategoryAndAction(ChangeLogTypeBuiltin.GROUP_DELETE)) {
//print the name from the entry
System.out.println("Group delete, name: "
+ changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_DELETE.name));
}
if (changeLogEntry.equalsCategoryAndAction(ChangeLogTypeBuiltin.GROUP_UPDATE)) {
//print the name from the entry
System.out.println("Group update, name: "
+ changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_UPDATE.name)
+ ", property: " + changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_UPDATE.propertyChanged)
+ ", from: '" + changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_UPDATE.propertyOldValue)
+ "', to: '" + changeLogEntry.retrieveValueForLabel(ChangeLogLabels.GROUP_UPDATE.propertyNewValue) + "'");
}
//we successfully processed this record
}
} catch (Exception e) {
changeLogProcessorMetadata.registerProblem(e, "Error processing record", currentId);
//we made it to this -1
return currentId-1;
}
if (currentId == -1) {
throw new RuntimeException("Couldnt process any records");
}
return currentId;
}
}
4. Register this consumer in the grouper-loader.properties (note, the cron string might not be needed. If blank, it will default to on the minute, every minute [right after the temp daemon]. If there are multiple consumers, they will be staggered by 2 seconds)
#specify the consumers here. specify the consumer name after the changeLog.consumer. part. This example is "ldappc" #but it could be changeLog.consumer.myConsumerName.class #the class must extend edu.internet2.middleware.grouper.changeLog.ChangeLogConsumerBase #changeLog.consumer.ldappc.class = #the quartz cron is a cron-like string. it defaults to every minute on the minute (since the temp to change log job runs #at 10 seconds to each minute). it defaults to this: 0 * * * * ? #though it will stagger each one by 2 seconds #changeLog.consumer.ldappc.quartzCron = changeLog.consumer.printTest.class = edu.internet2.middleware.grouper.changeLog.consumer.PrintTest changeLog.consumer.printTest.quartzCron =
5. To test this, run the loader, and a GSH shell in separate windows. Enter commands which add/delete types, and add/update/delete groups in the GSH shell, see print statements to STDOUT in the loader window. Start the loader:
C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper>bin\gsh -loader Using GROUPER_HOME: C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper\bin\.. Using GROUPER_CONF: C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper\bin\../conf Using JAVA: "c:\dev_inst\java/bin/java" using MEMORY: 64m-512m Grouper starting up: version: 1.5.0-rc1, build date: 2009/06/10 01:09:59, env: DEV ...
6. Start up GSH, and enter some commands
C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper>bin\gsh
Using GROUPER_HOME: C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper\bin\..
Using GROUPER_CONF: C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper\bin\../conf
Using JAVA: "c:\dev_inst\java/bin/java"
using MEMORY: 64m-512m
...
gsh 0% typeAdd("testA");
type: 'testA'
gsh 1% addRootStem("stem1", "stem1");
stem: name='stem1' displayName='stem1' uuid='ece24338-afde-4d5f-8593-d11b8b4341aa'
gsh 2% group = addGroup("stem1", "group1", "group1");
group: name='stem1:group1' displayName='stem1:group1' uuid='6ada2741-20b0-4cb6-9780-b07cfaf31db3'
gsh 3% typeDel("testA");
true
gsh 4% group.setDisplayExtension("group2");
gsh 5% group.setDescription("my description");
gsh 6% group.store();
gsh 7% typeAdd("testB");
type: 'testB'
gsh 8% typeDel("testB");
true
gsh 9% group.delete();
gsh 10%
7. Switch back to the loader window, and see the output (especially after waiting until the top of the minute, so that the daemon has run again). Note there are multiple entries for one group update since multiple properties have changed.
C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper>bin\gsh -loader Using GROUPER_HOME: C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper\bin\.. Using GROUPER_CONF: C:\mchyzer\isc\dev\grouper-qs-1.2.0\grouper\bin\../conf Using JAVA: "c:\dev_inst\java/bin/java" using MEMORY: 64m-512m Group type add, name: testA Group add, name: stem1:group1 Group type delete, name: testA Group update, name: stem1:group1, property: description, from: 'null', to: 'my description' Group update, name: stem1:group1, property: displayExtension, from: 'group1', to: 'group2' Group type add, name: testB Group type delete, name: testB Group delete, name: stem1:group1
8. Note that the state of this consumer is stored in the database. So if grouper is shutdown, or the loader is shutdown, and brought back up, it will start up where it left off. The name is the name configured in the grouper-loader.properties above
select * from grouper_change_log_consumer name last_sequence_processed last_updated created_on id hibernate_version_number -------------------------------------------------------------------------------------------------------------------------------- printTest 14 1244611320119 1244610780084 57bd90bc-64b9-43c7-8023-dbb2b5fcbd03 4
To do
- Add retry timeouts and max retries (currently it will just keep retrying every cycle of loader (every minute?) if there is an exception in target system. It should sleep for progressively longer and longer until hitting a max
- Add a friendly view so it is easy to look at records
- Add in other notification types: attributes, group rename, privileges, etc
- Make sure effecive memberships with the new membership db layout
- Add a cleanup daemon which deletes old or too many records (configurable)
- Add filters so that all systems are not notified on all change log entries
- Add a web service to read notifications
- Add hook on change log entry (pre/post insert/update/delete
- Add import/export to xml
Design / FAQ
- The change log is transactional. If a rollback or failure occurs in Grouper, the change log will be in sync.
- If the destination is unreachable, it will retry next time with the same record
- You can query the change log directly with the API through the DAO
- Change log entries are ordered so that they norify in an order that will work (i.e. create a group before adding members)
- There is a unique id (sequential sequence) for each row in change log
- The timestamp is in micros, and will never be the same on the same jvm as another record
- No database triggers are user, its an all Java solution
- There are 12 columns to stash data in the change log entry
- Grouper keeps track of progress of consumers run in the loader